I’m currently planning to build a new server after I discovered what my system uses in Idle. As I have to set up a new system anyways I would like to add a NAS to it to manige my storage. Currently I just have a zfs-Pool in proxmox for my Data-Drives and all VMs/Containers that need access have escalated rights and can directly access the pool (and all other storage on proxmox) which is a bit janky and definetly not best practice security-wise. Another negative side effect is that the drives are barely spun down. Thats why I now want to have a Nas as the only System controling the Drive pool. Here’s where my question comes up: Should I run TrueNas (scale?) in a VM and pass the drives through somehow (is that possible without mounting them in Proxmox, as I would like them fully controled by the Nas, including running the zfs pool, etc. ?)? Or do I install TrueNas scale and then run Proxmox as a VM inside, would my performance penalty be huge here, would I still be able to pass throught USB/PCI devices (maybe even the cpu’s igup to forward that to jellyfin if that’s even possible in Proxmox?)?

  • You999@lemmy.world
    link
    fedilink
    English
    arrow-up
    4
    ·
    1 year ago

    The best way to do this is to run a truenas VM within proxmox and passthrough your HBA into the truenas VM. That will give truenas full control over any drive connected to that HBA. The performance overhead isn’t that much so don’t worry about it.

    • dustojnikhummer@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      In a pinch passing through drives also works with ZFS, but obviously you lose SMART etc. I ran that for a few weeks before I managed to get ACS override working and my ZFS pool got picked up when TrueNAS started.

      • You999@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 year ago

        ZFS not have access to smart only works up until a drive starts acting up. Without SMART ZFS can’t accurately determine if a drive is failing and lock the pool in order to prevent further data loss.

        • dustojnikhummer@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          1 year ago

          I’m aware, but I’m saying in a pinch it will work and when you pass your HBA fully you won’t need to reconfigure anything, at least I didn’t

  • eleitl@lemmy.ml
    link
    fedilink
    English
    arrow-up
    2
    ·
    1 year ago

    You might want to look into all-in-one zfs, which exports your pool as an NFS share internally and externally. In case you use spinning rust, L2ARC and ZIL are your friends.

  • carzian@lemmy.ml
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    There have been many posts about people running truenas as a VM in proxmox. There are a few things to consider that I’m not well versed on, so I suggest doing some more in depth research, but it’s definitely possible (I did it myself up until the end of last year).

    One of the easiest ways to get the hard drives into truenas is to connect them to a raid card running in IT mode, which allows the OS to directly control the drives (do not raid them, truenas wants the raw disks), and then pass the raid card to the truenas VM

      • einsteinx2@programming.dev
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 year ago

        Oh my home NAS I started this way and it worked great. I only bought an HBA card because I needed more ports. Your mobo probably exposes your SATA controller as a PCI-E device that can be used via pass through in a VM. In my case I booted Proxmox off of NVME drive and passed my SATA controller to a Debian VM where I just use simple NFS and Samba for sharing and SnapRAID for drive parity (but TrueNas should work just as well).

        I had zero issues with it and when I upgraded to an HBA card I just switched the drives to those ports and switched the PCIE device I was passing through and everything just worked (helps I always mount using partition UUIDs).